grepmax 0.26.35 → 0.26.37

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
       "name": "grepmax",
       "source": "./plugins/grepmax",
       "description": "Semantic code search for Claude Code. Automatically indexes your project and provides intelligent search capabilities.",
-      "version": "0.26.35",
+      "version": "0.26.37",
       "author": {
         "name": "Robert Owens",
         "email": "robowens@me.com"

package/README.md

@@ -50,6 +50,7 @@ No setup required — gmax auto-detects your platform (GPU on Apple Silicon, CPU
 gmax setup                    # Interactive wizard (models, embedding mode, plugins)
 gmax config                   # View current settings
 gmax config --embed-mode gpu  # Switch to GPU (Apple Silicon)
+gmax config --worker-threads 2  # Cap worker processes (auto = default); applies on daemon restart
 gmax doctor                   # Health check
 gmax doctor --fix             # Auto-repair (compact, prune, remove stale locks)
 ```
@@ -85,6 +86,7 @@ gmax context src/lib/auth.ts --budget 4000 # Deterministic file/path context
 gmax project                  # Languages, structure, key symbols
 gmax related src/lib/auth.ts  # Dependencies + dependents
 gmax status                   # All indexed projects + chunk counts
+gmax status --json            # The same plus daemon, settings and workers, as JSON
 ```
 
 In our public benchmarks, `grepmax` can save about 20% of your LLM tokens and deliver a 30% speedup.
@@ -214,6 +216,7 @@ The daemon watches only projects that are in use: each Claude Code session (its
 ```bash
 gmax watch --daemon -b        # Start daemon manually
 gmax watch stop               # Stop daemon
+gmax watch restart            # Stop it, wait for it to exit, start a fresh one (--json)
 gmax status                   # See all projects + watcher status
 ```
 
@@ -423,7 +426,7 @@ fixtures/
 | Variable | Description | Default |
 | --- | --- | --- |
 | `GMAX_EMBED_MODE` | Force `cpu` or `gpu` | Auto-detect |
-| `GMAX_WORKER_THREADS` | Worker threads for embedding | 50% of cores |
+| `GMAX_WORKER_THREADS` | Worker processes for embedding; overrides `gmax config --worker-threads` | `min(4, max(2, cores/2))` |
 | `GMAX_WORKER_RSS_RECYCLE_MB` | Recycle workers that remain above this RSS; `0` disables the check | `1536` |
 | `GMAX_DEBUG` | Debug logging | Off |
 | `GMAX_SUMMARIZER` | Enable summarizer auto-start (`1`) | Off |
@@ -438,7 +441,7 @@ gmax doctor --fix             # Auto-repair (compact, prune, fix locks)
 gmax doctor --agent           # Machine-readable health output
 gmax index                    # Reindex (auto-detects and repairs cache/vector mismatches)
 gmax index --reset            # Full reindex from scratch
-gmax watch stop && gmax watch --daemon -b  # Restart daemon
+gmax watch restart            # Restart daemon
 ```
 
 `gmax doctor` reports ANN index state. `ANN: vector index not built` is normal with the default exact-search configuration.

package/dist/commands/config.js

@@ -1,4 +1,37 @@
 "use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
 var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
     function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
     return new (P || (P = Promise))(function (resolve, reject) {
@@ -10,6 +43,7 @@ var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, ge
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.config = void 0;
+const os = __importStar(require("node:os"));
 const commander_1 = require("commander");
 const config_1 = require("../config");
 const embedding_status_1 = require("../lib/index/embedding-status");
@@ -22,12 +56,15 @@ exports.config = new commander_1.Command("config")
     .option("--embed-mode <mode>", "Set embedding mode: cpu or gpu")
     .option("--model-tier <tier>", "Set model tier: small (384d) or standard (768d)")
     .option("--query-log <on|off>", "Enable/disable query logging to ~/.gmax/logs/queries.jsonl")
+    .option("--worker-threads <n|auto>", "Set the number of embedding worker processes, or auto for the default")
     .addHelpText("after", `
 Examples:
   gmax config                          Show current configuration
   gmax config --embed-mode cpu         Switch to CPU embeddings
   gmax config --model-tier standard    Switch to standard model (768d)
   gmax config --query-log on           Enable query logging
+  gmax config --worker-threads 2       Run at most 2 worker processes
+  gmax config --worker-threads auto    Back to the computed default
 `)
     .action((_opts, cmd) => __awaiter(void 0, void 0, void 0, function* () {
     var _a, _b, _c, _d, _e;
@@ -37,7 +74,8 @@ Examples:
     const project = (0, project_registry_1.getProject)((_a = (0, project_root_1.findProjectRoot)(process.cwd())) !== null && _a !== void 0 ? _a : process.cwd());
     const hasUpdates = options.embedMode !== undefined ||
         options.modelTier !== undefined ||
-        options.queryLog !== undefined;
+        options.queryLog !== undefined ||
+        options.workerThreads !== undefined;
     if (!hasUpdates) {
         // Show current config
         const tier = (_b = config_1.MODEL_TIERS[globalConfig.modelTier]) !== null && _b !== void 0 ? _b : config_1.MODEL_TIERS.small;
@@ -57,10 +95,15 @@ Examples:
             }
         }
         console.log(`  Query log:   ${globalConfig.queryLog ? "on" : "off"}`);
+        const threads = (0, config_1.resolveWorkerThreads)({
+            env: process.env.GMAX_WORKER_THREADS,
+            configValue: globalConfig.workerThreads,
+        });
+        console.log(`  Worker threads: ${threads.value} (${threads.source})`);
         if (project === null || project === void 0 ? void 0 : project.lastIndexed) {
             console.log(`  Last indexed: ${project.lastIndexed}`);
         }
-        console.log(`\nTo change: gmax config --embed-mode <cpu|gpu> --model-tier <small|standard> --query-log <on|off>`);
+        console.log(`\nTo change: gmax config --embed-mode <cpu|gpu> --model-tier <small|standard> --query-log <on|off> --worker-threads <n|auto>`);
         yield (0, exit_1.gracefulExit)();
         return;
     }
@@ -75,6 +118,36 @@ Examples:
         yield (0, exit_1.gracefulExit)(1);
         return;
     }
+    if (options.workerThreads !== undefined) {
+        const cores = os.cpus().length || 1;
+        const raw = options.workerThreads.trim().toLowerCase();
+        const value = raw === "auto" ? undefined : (0, config_1.parseWorkerThreads)(raw, cores);
+        if (value === null) {
+            console.error(`Invalid worker threads: ${options.workerThreads} (use a whole number from 1 to ${cores}, or auto)`);
+            yield (0, exit_1.gracefulExit)(1);
+            return;
+        }
+        (0, index_config_1.writeGlobalConfig)(Object.assign(Object.assign({}, globalConfig), { workerThreads: value }));
+        globalConfig.workerThreads = value;
+        const effective = (0, config_1.resolveWorkerThreads)({
+            env: process.env.GMAX_WORKER_THREADS,
+            configValue: value,
+            cores,
+        });
+        console.log(value === undefined
+            ? `Worker threads set to auto (${(0, config_1.defaultWorkerThreads)(cores)} on this machine).`
+            : `Worker threads set to ${value}.`);
+        if (effective.source === "env") {
+            console.log(`GMAX_WORKER_THREADS=${process.env.GMAX_WORKER_THREADS} is set in this shell and overrides it here.`);
+        }
+        console.log("Takes effect when the gmax daemon restarts: gmax watch stop, then gmax watch --daemon -b.");
+        if (options.queryLog === undefined &&
+            !options.embedMode &&
+            !options.modelTier) {
+            yield (0, exit_1.gracefulExit)();
+            return;
+        }
+    }
     // Handle query-log toggle (independent of model/embed changes)
     if (options.queryLog !== undefined) {
         if (!["on", "off"].includes(options.queryLog)) {
@@ -103,6 +176,7 @@ Examples:
         embedMode: newMode,
         mlxModel: newMode === "gpu" ? tier.mlxModel : undefined,
         queryLog: globalConfig.queryLog,
+        workerThreads: globalConfig.workerThreads,
     });
     (0, index_config_1.writeSetupConfig)(paths.configPath, {
         embedMode: newMode,

package/dist/commands/status.js

@@ -43,6 +43,8 @@ var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, ge
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.status = void 0;
+exports.buildStatusJson = buildStatusJson;
+const node_child_process_1 = require("node:child_process");
 const os = __importStar(require("node:os"));
 const commander_1 = require("commander");
 const config_1 = require("../config");
@@ -89,7 +91,17 @@ function loadStatusView(projects) {
                         chunkCounts.set(project.root, stats.chunks);
                     }
                 }
-                return { watchers, chunkCounts };
+                const daemon = {
+                    pid: typeof resp.pid === "number" ? resp.pid : null,
+                    uptimeSec: typeof resp.uptime === "number" ? resp.uptime : null,
+                    // Older daemons do not report these.
+                    workers: typeof resp.workers === "number" ? resp.workers : null,
+                    workerThreads: resp.workerThreads &&
+                        typeof resp.workerThreads.value === "number"
+                        ? resp.workerThreads.value
+                        : null,
+                };
+                return { watchers, chunkCounts, daemon };
             }),
             inProcess: () => __awaiter(this, void 0, void 0, function* () {
                 (0, watcher_store_1.listWatchers)(); // cleans stale entries as side effect
@@ -129,6 +141,86 @@ function loadStatusView(projects) {
         });
     });
 }
+/** State shown per project; shared by --agent and --json. */
+function projectState(project, watcher) {
+    var _a;
+    const projectStatus = (_a = project.status) !== null && _a !== void 0 ? _a : "indexed";
+    if (projectStatus === "pending")
+        return "pending";
+    if (projectStatus === "error")
+        return "error";
+    if ((watcher === null || watcher === void 0 ? void 0 : watcher.status) === "syncing")
+        return "indexing";
+    if (watcher)
+        return "watching";
+    return "idle";
+}
+/**
+ * Worker processes under the daemon, from the process table. Used only when
+ * the daemon is too old to report its pool size.
+ */
+function countWorkerProcesses(daemonPid) {
+    try {
+        const out = (0, node_child_process_1.execFileSync)("ps", ["-axo", "ppid=,command="], {
+            encoding: "utf8",
+            timeout: 2000,
+        });
+        let n = 0;
+        for (const line of out.split("\n")) {
+            const m = line.trim().match(/^(\d+)\s+(\S+)/);
+            if (m && Number(m[1]) === daemonPid && m[2] === "gmax-worker")
+                n++;
+        }
+        return n;
+    }
+    catch (_a) {
+        return null;
+    }
+}
+function toMs(iso) {
+    if (!iso)
+        return null;
+    const t = Date.parse(iso);
+    return Number.isFinite(t) ? t : null;
+}
+function buildStatusJson(input) {
+    var _a, _b, _c, _d, _e;
+    const now = (_a = input.now) !== null && _a !== void 0 ? _a : Date.now();
+    const { view, globalConfig } = input;
+    const d = view.daemon;
+    let workersRunning = null;
+    if (d) {
+        workersRunning = (_b = d.workers) !== null && _b !== void 0 ? _b : (d.pid !== null
+            ? ((_c = input.countWorkers) !== null && _c !== void 0 ? _c : countWorkerProcesses)(d.pid)
+            : null);
+    }
+    return {
+        daemon: {
+            running: d !== undefined,
+            pid: (_d = d === null || d === void 0 ? void 0 : d.pid) !== null && _d !== void 0 ? _d : null,
+            since: (d === null || d === void 0 ? void 0 : d.uptimeSec) != null ? now - d.uptimeSec * 1000 : null,
+            workerThreads: (_e = d === null || d === void 0 ? void 0 : d.workerThreads) !== null && _e !== void 0 ? _e : null,
+        },
+        settings: {
+            embedMode: globalConfig.embedMode,
+            modelTier: globalConfig.modelTier,
+            vectorDim: globalConfig.vectorDim,
+            queryLog: globalConfig.queryLog === true,
+            workerThreads: input.workerThreads,
+        },
+        workersRunning,
+        indexing: input.indexing,
+        projects: input.projects.map((project) => { var _a, _b; return ({
+            name: project.name,
+            root: project.root,
+            chunks: (_b = (_a = view.chunkCounts.get(project.root)) !== null && _a !== void 0 ? _a : project.chunkCount) !== null && _b !== void 0 ? _b : 0,
+            indexedAt: toMs(project.lastIndexed),
+            state: projectState(project, view.watchers.get(project.root)),
+            embedding: (0, embedding_status_1.projectEmbeddingStatus)(project, globalConfig).state,
+        }); }),
+        at: now,
+    };
+}
 const style = {
     bold: (s) => `\x1b[1m${s}\x1b[22m`,
     dim: (s) => `\x1b[2m${s}\x1b[22m`,
@@ -169,12 +261,14 @@ function formatChunks(n) {
 exports.status = new commander_1.Command("status")
     .description("Show gmax index status for all projects")
     .option("--agent", "Compact output for AI agents", false)
+    .option("--json", "Machine-readable status: daemon, settings, projects", false)
     .addHelpText("after", `
 Examples:
   gmax status              Show status of all indexed projects
+  gmax status --json       The same, plus daemon and settings, as JSON
 `)
     .action((opts) => __awaiter(void 0, void 0, void 0, function* () {
-    var _a, _b, _c, _d;
+    var _a, _b, _c;
     const globalConfig = (0, index_config_1.readGlobalConfig)();
     const projects = (0, project_registry_1.listProjects)();
     const indexing = (0, lock_1.isLocked)(config_1.PATHS.globalRoot);
@@ -192,6 +286,21 @@ Examples:
         throw err;
     }
     const { watchers, chunkCounts } = view;
+    if (opts.json) {
+        const json = buildStatusJson({
+            view,
+            projects,
+            globalConfig,
+            indexing,
+            workerThreads: (0, config_1.resolveWorkerThreads)({
+                env: process.env.GMAX_WORKER_THREADS,
+                configValue: globalConfig.workerThreads,
+            }),
+        });
+        console.log(JSON.stringify(json));
+        yield (0, exit_1.gracefulExit)();
+        return;
+    }
     if (!opts.agent) {
         // Header
         console.log(`\n${style.bold("gmax")} · ${globalConfig.modelTier} (${globalConfig.vectorDim}d, ${globalConfig.embedMode})${indexing ? style.yellow(" · indexing...") : ""}`);
@@ -208,21 +317,9 @@ Examples:
     }
     if (opts.agent) {
         for (const project of projects) {
-            const watcher = watchers.get(project.root);
-            const projectStatus = (_a = project.status) !== null && _a !== void 0 ? _a : "indexed";
-            let st;
-            if (projectStatus === "pending")
-                st = "pending";
-            else if (projectStatus === "error")
-                st = "error";
-            else if ((watcher === null || watcher === void 0 ? void 0 : watcher.status) === "syncing")
-                st = "indexing";
-            else if (watcher)
-                st = "watching";
-            else
-                st = "idle";
+            const st = projectState(project, watchers.get(project.root));
             const isCurrent = project.root === currentRoot;
-            const count = (_b = chunkCounts.get(project.root)) !== null && _b !== void 0 ? _b : project.chunkCount;
+            const count = (_a = chunkCounts.get(project.root)) !== null && _a !== void 0 ? _a : project.chunkCount;
             const identity = (0, embedding_status_1.projectEmbeddingStatus)(project, globalConfig);
             console.log(`${project.name}\t${formatChunks(count)}\t${formatAge(project.lastIndexed)}\t${st}\tembedding=${identity.state}${isCurrent ? "\tcurrent" : ""}`);
         }
@@ -240,7 +337,7 @@ Examples:
         const watcher = watchers.get(project.root);
         // Status column
         let statusStr;
-        const projectStatus = (_c = project.status) !== null && _c !== void 0 ? _c : "indexed";
+        const projectStatus = (_b = project.status) !== null && _b !== void 0 ? _b : "indexed";
         if (projectStatus === "pending") {
             statusStr = style.yellow("pending");
         }
@@ -257,7 +354,7 @@ Examples:
             statusStr = style.dim("idle");
         }
         // Chunks column
-        const count = (_d = chunkCounts.get(project.root)) !== null && _d !== void 0 ? _d : project.chunkCount;
+        const count = (_c = chunkCounts.get(project.root)) !== null && _c !== void 0 ? _c : project.chunkCount;
         const chunks = `${formatChunks(count)} chunks`;
         // Age column
         const age = formatAge(project.lastIndexed);

package/dist/commands/watch.js

@@ -52,6 +52,7 @@ const syncer_1 = require("../lib/index/syncer");
 const watcher_1 = require("../lib/index/watcher");
 const meta_cache_1 = require("../lib/store/meta-cache");
 const vector_db_1 = require("../lib/store/vector-db");
+const daemon_restart_1 = require("../lib/utils/daemon-restart");
 const exit_1 = require("../lib/utils/exit");
 const filter_builder_1 = require("../lib/utils/filter-builder");
 const log_rotate_1 = require("../lib/utils/log-rotate");
@@ -131,20 +132,10 @@ exports.watch = new commander_1.Command("watch")
                 // competitor; the startup code would only kill the busy peer.
                 process.exit(0);
             }
-            const logFile = path.join(config_1.PATHS.logsDir, "daemon.log");
-            const out = (0, log_rotate_1.openRotatedLog)(logFile);
-            let child;
+            const { spawnDaemonProcess } = yield Promise.resolve().then(() => __importStar(require("../lib/utils/daemon-launcher")));
+            let started;
             try {
-                child = (0, node_child_process_1.spawn)(process.argv[0], [process.argv[1], "watch", "--daemon"], {
-                    detached: true,
-                    stdio: ["ignore", out, out],
-                    cwd: process.cwd(),
-                    env: Object.assign(Object.assign({}, process.env), { GMAX_BACKGROUND: "true" }),
-                });
-                yield new Promise((resolve, reject) => {
-                    child.once("spawn", resolve);
-                    child.once("error", reject);
-                });
+                started = yield spawnDaemonProcess();
             }
             catch (error) {
                 const message = error instanceof Error ? error.message : String(error);
@@ -152,14 +143,7 @@ exports.watch = new commander_1.Command("watch")
                 process.exitCode = 1;
                 return;
             }
-            finally {
-                try {
-                    fs.closeSync(out);
-                }
-                catch (_c) { }
-            }
-            child.unref();
-            console.log(`Daemon started (PID: ${child.pid}, log: ${logFile})`);
+            console.log(`Daemon started (PID: ${started.pid}, log: ${started.logFile})`);
             process.exit(0);
         }
         // Daemon foreground
@@ -210,7 +194,7 @@ exports.watch = new commander_1.Command("watch")
             try {
                 yield daemon.shutdown();
             }
-            catch (_d) { }
+            catch (_c) { }
             process.exitCode = 1;
             return;
         }
@@ -258,7 +242,7 @@ exports.watch = new commander_1.Command("watch")
             try {
                 fs.closeSync(out);
             }
-            catch (_e) { }
+            catch (_d) { }
         }
         child.unref();
         console.log(`Watcher started for ${projectName} (PID: ${child.pid}, log: ${logFile})`);
@@ -413,6 +397,85 @@ exports.watch
     }
     yield (0, exit_1.gracefulExit)();
 }));
+exports.watch
+    .command("restart")
+    .description("Stop the daemon and start a fresh one; waits for the old one to exit")
+    .option("--json", "Machine-readable result", false)
+    .action((options) => __awaiter(void 0, void 0, void 0, function* () {
+    const emit = (result) => {
+        if (options.json)
+            console.log(JSON.stringify(result));
+        else if (result.ok)
+            console.log((0, daemon_restart_1.formatRestartResult)(result));
+        else
+            console.error((0, daemon_restart_1.formatRestartResult)(result));
+    };
+    const refuse = (error) => {
+        emit({
+            ok: false,
+            previousPid: null,
+            stoppedVia: null,
+            pid: null,
+            version: null,
+            workers: null,
+            workerThreads: null,
+            adopted: false,
+            readyMs: null,
+            error,
+        });
+    };
+    // Refused before anything is stopped: a daemon that cannot be started
+    // again from here must not be taken down.
+    if (process.env.GMAX_SECONDARY_STORE === "1") {
+        refuse("this project's index lives on an external drive and never runs a daemon");
+        yield (0, exit_1.gracefulExit)(2);
+        return;
+    }
+    const { storeWriteDeniedNotice } = yield Promise.resolve().then(() => __importStar(require("../lib/utils/store-access")));
+    const denied = storeWriteDeniedNotice();
+    if (denied) {
+        refuse(denied);
+        yield (0, exit_1.gracefulExit)(2);
+        return;
+    }
+    const client = yield Promise.resolve().then(() => __importStar(require("../lib/utils/daemon-client")));
+    const { spawnDaemonProcess } = yield Promise.resolve().then(() => __importStar(require("../lib/utils/daemon-launcher")));
+    let parentCmd = "?";
+    try {
+        const { execSync } = yield Promise.resolve().then(() => __importStar(require("node:child_process")));
+        parentCmd = execSync(`ps -o command= -p ${process.ppid}`, {
+            encoding: "utf8",
+        }).trim();
+    }
+    catch (_a) { }
+    const result = yield (0, daemon_restart_1.restartDaemon)({
+        ping: (timeoutMs) => client.sendDaemonCommand({ cmd: "ping" }, { timeoutMs }),
+        status: () => client.sendDaemonCommand({ cmd: "status" }),
+        requestShutdown: () => client.sendDaemonCommand({
+            cmd: "shutdown",
+            reason: "gmax-watch-restart",
+            from_pid: process.pid,
+            from_ppid: process.ppid,
+            from_argv: process.argv.slice(0, 4),
+            from_parent_cmd: parentCmd,
+        }, { timeoutMs: 10000 }),
+        liveDaemonPidFromFiles: () => client.isDaemonHeartbeatFresh() ? client.readDaemonPid() : null,
+        isAlive: watcher_store_1.isProcessRunning,
+        terminate: (pid) => process.kill(pid, "SIGTERM"),
+        spawn: () => __awaiter(void 0, void 0, void 0, function* () {
+            try {
+                return (yield spawnDaemonProcess()).pid;
+            }
+            catch (_a) {
+                return null;
+            }
+        }),
+        sleep: (ms) => new Promise((r) => setTimeout(r, ms)),
+        now: () => Date.now(),
+    });
+    emit(result);
+    yield (0, exit_1.gracefulExit)(result.ok ? 0 : 1);
+}));
 exports.watch
     .command("stop")
     .description("Stop watcher for current project")

package/dist/config.js

@@ -33,11 +33,15 @@ var __importStar = (this && this.__importStar) || (function () {
     };
 })();
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.INDEXABLE_EXTENSIONS = exports.STALE_TEMP_FILE_AGE_MS = exports.VERSION_PRUNE_THRESHOLD = exports.FRAGMENT_COMPACT_THRESHOLD = exports.DISK_LOW_BYTES = exports.DISK_CRITICAL_BYTES = exports.MAX_FILE_SIZE_BYTES = exports.PATHS = exports.MAX_WORKER_MEMORY_MB = exports.WORKER_BOOT_TIMEOUT_MS = exports.WORKER_TIMEOUT_MS = exports.REBUILD_COMMAND = exports.CHUNKER_VERSION_HISTORY = exports.CONFIG = exports.MODEL_IDS = exports.DEFAULT_MODEL_TIER = exports.MODEL_TIERS = void 0;
+exports.INDEXABLE_EXTENSIONS = exports.STALE_TEMP_FILE_AGE_MS = exports.VERSION_PRUNE_THRESHOLD = exports.FRAGMENT_COMPACT_THRESHOLD = exports.DISK_LOW_BYTES = exports.DISK_CRITICAL_BYTES = exports.MAX_FILE_SIZE_BYTES = exports.PATHS = exports.MAX_WORKER_MEMORY_MB = exports.WORKER_BOOT_TIMEOUT_MS = exports.WORKER_TIMEOUT_MS = exports.REBUILD_COMMAND = exports.CHUNKER_VERSION_HISTORY = exports.CONFIG = exports.WORKER_THREADS_SETTING = exports.MODEL_IDS = exports.DEFAULT_MODEL_TIER = exports.MODEL_TIERS = void 0;
+exports.defaultWorkerThreads = defaultWorkerThreads;
+exports.parseWorkerThreads = parseWorkerThreads;
+exports.resolveWorkerThreads = resolveWorkerThreads;
 exports.describeChunkerGap = describeChunkerGap;
 exports.describeEmbeddingGap = describeEmbeddingGap;
 exports.describeSchemaDimGap = describeSchemaDimGap;
 exports.schemaDimAgentRow = schemaDimAgentRow;
+const fs = __importStar(require("node:fs"));
 const os = __importStar(require("node:os"));
 const path = __importStar(require("node:path"));
 exports.MODEL_TIERS = {
@@ -63,27 +67,79 @@ exports.MODEL_IDS = {
     embed: exports.MODEL_TIERS[exports.DEFAULT_MODEL_TIER].onnxModel,
     colbert: "ryandono/mxbai-edge-colbert-v0-17m-onnx-int8",
 };
-const DEFAULT_WORKER_THREADS = (() => {
-    var _a;
-    const fromEnv = Number.parseInt((_a = process.env.GMAX_WORKER_THREADS) !== null && _a !== void 0 ? _a : "", 10);
-    if (Number.isFinite(fromEnv) && fromEnv > 0)
-        return fromEnv;
-    // Each worker is a separate process that loads the ONNX models, peaking at
-    // ~800 MB-1 GB. The old max(4, cores/2) let a 14-core host run 7 of them;
-    // together with the daemon that froze a 48 GB machine. Four is the ceiling;
-    // two is the floor where the cores exist, so one worker can stay free for
-    // search. GMAX_WORKER_THREADS still overrides.
-    const cores = os.cpus().length || 1;
+const HOME = os.homedir();
+// Models, grammars and the pinned HF cache are shared by every store and always
+// live on the internal disk. GMAX_HOME moves only the data root: a secondary
+// store on an external drive (see src/lib/utils/stores.ts) sets it before this
+// module loads, so a project's index can live on the drive with its files.
+const SHARED_ROOT = path.join(HOME, ".gmax");
+const GLOBAL_ROOT = process.env.GMAX_HOME
+    ? path.resolve(process.env.GMAX_HOME)
+    : SHARED_ROOT;
+function coreCount() {
+    return os.cpus().length || 1;
+}
+/**
+ * Each worker is a separate process that loads the ONNX models, peaking at
+ * ~800 MB-1 GB. The old max(4, cores/2) let a 14-core host run 7 of them;
+ * together with the daemon that froze a 48 GB machine. Four is the ceiling;
+ * two is the floor where the cores exist, so one worker can stay free for
+ * search.
+ */
+function defaultWorkerThreads(cores = coreCount()) {
     const target = Math.min(4, Math.max(2, Math.floor(cores * 0.5)));
     return Math.max(1, Math.min(target, cores));
-})();
+}
+/** A stored or typed worker count is a whole number from 1 to the core count. */
+function parseWorkerThreads(value, cores = coreCount()) {
+    const n = typeof value === "number"
+        ? value
+        : typeof value === "string" && /^\d+$/.test(value.trim())
+            ? Number(value.trim())
+            : Number.NaN;
+    if (!Number.isInteger(n) || n < 1 || n > cores)
+        return null;
+    return n;
+}
+/**
+ * GMAX_WORKER_THREADS wins, then `workerThreads` in ~/.gmax/config.json
+ * (`gmax config --worker-threads`), then the computed default. The env var
+ * keeps its old meaning (any positive whole number); a stored value outside
+ * 1..cores is ignored.
+ */
+function resolveWorkerThreads(opts) {
+    var _a, _b;
+    const cores = (_a = opts.cores) !== null && _a !== void 0 ? _a : coreCount();
+    const fromEnv = Number.parseInt((_b = opts.env) !== null && _b !== void 0 ? _b : "", 10);
+    if (Number.isFinite(fromEnv) && fromEnv > 0) {
+        return { value: fromEnv, source: "env" };
+    }
+    const fromConfig = parseWorkerThreads(opts.configValue, cores);
+    if (fromConfig !== null)
+        return { value: fromConfig, source: "config" };
+    return { value: defaultWorkerThreads(cores), source: "default" };
+}
+function storedWorkerThreads() {
+    try {
+        const raw = fs.readFileSync(path.join(GLOBAL_ROOT, "config.json"), "utf-8");
+        return JSON.parse(raw).workerThreads;
+    }
+    catch (_a) {
+        return undefined;
+    }
+}
+/** Read once at load: a running daemon keeps the value it started with. */
+exports.WORKER_THREADS_SETTING = resolveWorkerThreads({
+    env: process.env.GMAX_WORKER_THREADS,
+    configValue: storedWorkerThreads(),
+});
 exports.CONFIG = {
     VECTOR_DIM: 384,
     COLBERT_DIM: 48,
     MAX_CHUNK_CHARS: 2000,
     MAX_CHUNK_LINES: 75,
     EMBED_BATCH_SIZE: 24,
-    WORKER_THREADS: DEFAULT_WORKER_THREADS,
+    WORKER_THREADS: exports.WORKER_THREADS_SETTING.value,
     QUERY_PREFIX: "",
     // Bump when chunk metadata semantics change in a way that requires a full
     // reindex to take effect. Must equal the latest entry's `v` in
@@ -217,15 +273,6 @@ exports.MAX_WORKER_MEMORY_MB = (() => {
     const HARD_CEILING = 4096; // 4GB max per worker regardless of system RAM
     return Math.min(HARD_CEILING, Math.max(2048, Math.floor((os.totalmem() / 1024 / 1024) * 0.5)));
 })();
-const HOME = os.homedir();
-// Models, grammars and the pinned HF cache are shared by every store and always
-// live on the internal disk. GMAX_HOME moves only the data root: a secondary
-// store on an external drive (see src/lib/utils/stores.ts) sets it before this
-// module loads, so a project's index can live on the drive with its files.
-const SHARED_ROOT = path.join(HOME, ".gmax");
-const GLOBAL_ROOT = process.env.GMAX_HOME
-    ? path.resolve(process.env.GMAX_HOME)
-    : SHARED_ROOT;
 exports.PATHS = {
     globalRoot: GLOBAL_ROOT,
     sharedRoot: SHARED_ROOT,

package/dist/lib/daemon/daemon.js

@@ -844,6 +844,12 @@ class Daemon {
     uptime() {
         return Math.floor((Date.now() - this.startTime) / 1000);
     }
+    /** Worker processes alive in the pool right now (0 before it exists). */
+    workerCount() {
+        var _a;
+        var _b;
+        return (_b = (_a = this.workerPool) === null || _a === void 0 ? void 0 : _a.getWorkerPids().length) !== null && _b !== void 0 ? _b : 0;
+    }
     /** True once shared resources (LanceDB + MetaCache) are open. */
     isReady() {
         return this.ready;

package/dist/lib/daemon/ipc-handler.js

@@ -49,6 +49,7 @@ exports.startHeartbeat = startHeartbeat;
 exports.handleCommand = handleCommand;
 const fs = __importStar(require("node:fs"));
 const path = __importStar(require("node:path"));
+const config_1 = require("../../config");
 const logger_1 = require("../utils/logger");
 const operation_coordinator_1 = require("../utils/operation-coordinator");
 const path_containment_1 = require("../utils/path-containment");
@@ -183,6 +184,8 @@ function handleCommand(daemon, cmd, conn) {
                         leases: daemon.listWatchLeases(),
                         diskPressure: daemon.getDiskPressure(),
                         mlx: daemon.getMlxStatus(),
+                        workers: daemon.workerCount(),
+                        workerThreads: config_1.WORKER_THREADS_SETTING,
                     };
                 case "optimize":
                     // Keeps compaction single-writer: see Daemon.runOptimize.

package/dist/lib/utils/daemon-launcher.js

@@ -43,6 +43,7 @@ var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, ge
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.spawnDaemon = spawnDaemon;
+exports.spawnDaemonProcess = spawnDaemonProcess;
 const node_child_process_1 = require("node:child_process");
 const fs = __importStar(require("node:fs"));
 const path = __importStar(require("node:path"));
@@ -82,3 +83,36 @@ function spawnDaemon() {
         }
     });
 }
+/**
+ * Spawn the daemon itself (`watch --daemon`, not the `-b` launcher), detached,
+ * logging to daemon.log. Resolves to the daemon's own PID; throws when the
+ * spawn fails.
+ */
+function spawnDaemonProcess() {
+    return __awaiter(this, void 0, void 0, function* () {
+        const logFile = path.join(config_1.PATHS.logsDir, "daemon.log");
+        const out = (0, log_rotate_1.openRotatedLog)(logFile);
+        try {
+            const child = (0, node_child_process_1.spawn)(process.argv[0], [process.argv[1], "watch", "--daemon"], {
+                detached: true,
+                stdio: ["ignore", out, out],
+                cwd: process.cwd(),
+                env: Object.assign(Object.assign({}, process.env), { GMAX_BACKGROUND: "true" }),
+            });
+            yield new Promise((resolve, reject) => {
+                child.once("spawn", resolve);
+                child.once("error", reject);
+            });
+            child.unref();
+            if (child.pid === undefined)
+                throw new Error("spawned without a PID");
+            return { pid: child.pid, logFile };
+        }
+        finally {
+            try {
+                fs.closeSync(out);
+            }
+            catch (_a) { }
+        }
+    });
+}

package/dist/lib/utils/daemon-restart.js

@@ -0,0 +1,151 @@
+"use strict";
+var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
+    function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
+    return new (P || (P = Promise))(function (resolve, reject) {
+        function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }
+        function rejected(value) { try { step(generator["throw"](value)); } catch (e) { reject(e); } }
+        function step(result) { result.done ? resolve(result.value) : adopt(result.value).then(fulfilled, rejected); }
+        step((generator = generator.apply(thisArg, _arguments || [])).next());
+    });
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.restartDaemon = restartDaemon;
+exports.formatRestartResult = formatRestartResult;
+// Matches the daemon's own drain grace (DRAIN_GRACE_MS in daemon-client): a
+// graceful shutdown can legitimately run this long on a large index.
+const DEFAULT_EXIT_TIMEOUT_MS = 90000;
+const DEFAULT_READY_TIMEOUT_MS = 60000;
+const DEFAULT_POLL_MS = 200;
+function pidOf(resp) {
+    return typeof resp.pid === "number" && resp.pid > 0 ? resp.pid : null;
+}
+function restartDaemon(deps_1) {
+    return __awaiter(this, arguments, void 0, function* (deps, opts = {}) {
+        var _a, _b, _c, _d;
+        const exitTimeoutMs = (_a = opts.exitTimeoutMs) !== null && _a !== void 0 ? _a : DEFAULT_EXIT_TIMEOUT_MS;
+        const readyTimeoutMs = (_b = opts.readyTimeoutMs) !== null && _b !== void 0 ? _b : DEFAULT_READY_TIMEOUT_MS;
+        const pollMs = (_c = opts.pollMs) !== null && _c !== void 0 ? _c : DEFAULT_POLL_MS;
+        const result = {
+            ok: false,
+            previousPid: null,
+            stoppedVia: null,
+            pid: null,
+            version: null,
+            workers: null,
+            workerThreads: null,
+            adopted: false,
+            readyMs: null,
+        };
+        // A busy daemon can miss a short ping while its heartbeat is still fresh;
+        // the files catch that case so it is stopped rather than raced.
+        const first = yield deps.ping(5000);
+        const previousPid = first.ok
+            ? ((_d = pidOf(first)) !== null && _d !== void 0 ? _d : deps.liveDaemonPidFromFiles())
+            : deps.liveDaemonPidFromFiles();
+        result.previousPid = previousPid;
+        if (first.ok && previousPid === null) {
+            result.error =
+                "a daemon answered but did not name its PID; stop it with `gmax watch stop` first";
+            return result;
+        }
+        if (previousPid !== null) {
+            const shutdown = yield deps.requestShutdown();
+            if (shutdown.ok) {
+                result.stoppedVia = "ipc";
+            }
+            else if (deps.isAlive(previousPid)) {
+                // SIGTERM runs the same graceful shutdown as the IPC verb.
+                try {
+                    deps.terminate(previousPid);
+                    result.stoppedVia = "sigterm";
+                }
+                catch (_e) { }
+            }
+            const deadline = deps.now() + exitTimeoutMs;
+            while (deps.isAlive(previousPid)) {
+                if (deps.now() >= deadline) {
+                    result.error = `the daemon (PID ${previousPid}) did not exit within ${Math.round(exitTimeoutMs / 1000)}s; it is still running and no second daemon was started`;
+                    return result;
+                }
+                yield deps.sleep(pollMs);
+            }
+        }
+        const startedAt = deps.now();
+        // Something else (a session hook, an MCP server) may have started a daemon
+        // in the gap. Spawning beside it would only lose the lock race, so use it.
+        const gap = yield deps.ping(2000);
+        let spawnedPid = null;
+        if (gap.ok && pidOf(gap) !== previousPid) {
+            result.adopted = true;
+        }
+        else {
+            spawnedPid = yield deps.spawn();
+            if (spawnedPid === null) {
+                result.error = "the new daemon could not be spawned";
+                return result;
+            }
+        }
+        const readyBy = startedAt + readyTimeoutMs;
+        for (;;) {
+            const resp = yield deps.ping(2000);
+            const pid = pidOf(resp);
+            if (resp.ok && pid !== null && pid !== previousPid) {
+                if (resp.ready !== false) {
+                    result.pid = pid;
+                    result.version = typeof resp.version === "string" ? resp.version : null;
+                    if (spawnedPid !== null && pid !== spawnedPid)
+                        result.adopted = true;
+                    break;
+                }
+            }
+            else if (spawnedPid !== null && !resp.ok && !deps.isAlive(spawnedPid)) {
+                result.error = `the new daemon (PID ${spawnedPid}) exited during startup; see ~/.gmax/logs/daemon.log`;
+                return result;
+            }
+            if (deps.now() >= readyBy) {
+                result.error = `the new daemon did not become ready within ${Math.round(readyTimeoutMs / 1000)}s${spawnedPid !== null ? ` (PID ${spawnedPid})` : ""}`;
+                result.pid = spawnedPid;
+                return result;
+            }
+            yield deps.sleep(pollMs);
+        }
+        result.readyMs = deps.now() - startedAt;
+        const status = yield deps.status();
+        if (status.ok) {
+            if (typeof status.workers === "number")
+                result.workers = status.workers;
+            const threads = status.workerThreads;
+            if (threads && typeof threads.value === "number") {
+                result.workerThreads = threads.value;
+            }
+        }
+        result.ok = true;
+        return result;
+    });
+}
+function formatRestartResult(r) {
+    var _a;
+    const lines = [];
+    if (r.previousPid !== null) {
+        lines.push(`Stopped daemon (PID ${r.previousPid}).`);
+    }
+    else if (r.ok || r.pid !== null) {
+        lines.push("No daemon was running.");
+    }
+    if (!r.ok) {
+        lines.push(`Restart failed: ${(_a = r.error) !== null && _a !== void 0 ? _a : "unknown error"}.`);
+        return lines.join("\n");
+    }
+    const parts = [`PID ${r.pid}`];
+    if (r.version)
+        parts.push(`v${r.version}`);
+    if (r.workerThreads !== null)
+        parts.push(`${r.workerThreads} worker threads`);
+    if (r.workers !== null) {
+        parts.push(`${r.workers} worker${r.workers === 1 ? "" : "s"} running`);
+    }
+    if (r.readyMs !== null)
+        parts.push(`ready in ${(r.readyMs / 1000).toFixed(1)}s`);
+    lines.push(`${r.adopted ? "Daemon already started by another caller" : "Daemon started"} (${parts.join(", ")}).`);
+    return lines.join("\n");
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "grepmax",
-  "version": "0.26.35",
+  "version": "0.26.37",
   "author": "Robert Owens <78518764+reowens@users.noreply.github.com>",
   "homepage": "https://github.com/reowens/grepmax",
   "bugs": {

package/plugins/grepmax/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "grepmax",
-  "version": "0.26.35",
+  "version": "0.26.37",
   "description": "Semantic code search for Claude Code. Automatically indexes your project and provides intelligent search capabilities.",
   "author": {
     "name": "Robert Owens",